Soham Gugale

WORK HISTORY

EDUCATION

ABOUT SOHAM GUGALE

Most engineers pick one lane. I\'m building across three.I\'m a computational MEng student at Duke (Pratt School of Engineering), with a BTech in Civil Engineering from IIT Bombay (top 2% JEE Advanced, IMO regional qualifier). My work lives at the intersection of quantitative finance, machine learning, and high-performance systems — where the math is hard, and the stakes are real.What I\'ve built and shipped:A Physics-Informed Neural Network that solves the Black-Scholes PDE for European options — pricing at 0.09ms per valuation vs.~100ms for Finite Difference Methods, with sub-1% median error across test scenarios spanning volatility regimes and interest rate ranges.A Soft Actor-Critic market-making agent trained across RL timesteps on a custom Gymnasium LOB simulator steps/sec), achieving mean episode PnL of $35 over 500 episodes — deployed live via FastAPI + Docker on Render.A portfolio optimization engine using Hierarchical Risk Parity and Ledoit-Wolf shrinkage, achieving a 0.94 Sharpe Ratio in rolling walk-forward backtests (2018–2024) while cutting turnover by 63% vs. constrained Mean-Variance Optimization.FinFlow — a full-stack finance dashboard (React + FastAPI + PostgreSQL) with JWT auth, anomaly detection via z-score, spend forecasting with linear regression, and a live CI/CD pipeline on Vercel + Render.I think in systems. Whether it\'s a trading strategy, a neural PDE solver, or a production API, I care about correctness, speed, and robustness — not just getting something to run.Currently seeking summer 2025 internships in quant finance, ML engineering, or data science. Open to connect — especially if you work in quant, ML infrastructure, or computational finance.